Investigation of plate heat exchanger production data by machine learning methods, estimation and classification of lifetime test data

In this study, the measurement and test results of the Manisa Bosch Termoteknik factory in the production of plate heat exchangers in 2018 and 2019 were analyzed using Python programming language and machine learning methods. Heat exchanger plates produced under serial production conditions are transmitted to the quality control unit in certain periods and various measurements are made. During the operation of plate heat exchangers for various reasons, sudden pressure fluctuations occur. This behavior is called water hammer effect and is a criterion that plays a decisive role in the life of the plate heat exchanger. Bosch Termoteknik performs lifetime tests to observe the water hammer effect. The plate heat exchangers connected to the test stand are applied with pressure that will create a water hammer effect similar to the operating conditions of the boiler and the life values are determined. However, this test takes about 12 days and is very long in today's market and mass production conditions. The aim of this study is to accelerate the decision making process, to ensure quality and to increase efficiency based on past data. The data set consisting of measurement values and test results was analyzed using regression and classification methods. Numerical estimates were carried out with the linear regression method. In addition, classification analyzes were carried out with logistic regression, support vector machines, random forest and nearest neighbor algorithms, and the performances of the mathematical models created were compared.